1062 OPTICS LETTERS / Vol. 14, No.

19 / October 1, 1989

Theoretical modeling of erbium-doped fiber amplifiers with

excited-state absorption

P. R. Morkel and R. I. Laming

Optical Fibre Group, Department of Electronics and Computer Science, University of Southampton,
Highfield, Southampton SO9 5NH, UK

Received March 14, 1989; accepted June 30, 1989

Numerical solution of a general rate-equation model of an erbium-doped fiber amplifier highlights
several impor-
tant features of the device. A difference in strong-signal gain between copropagating and
counterpropagating
signal pump schemes in the presence of pump excited-state absorption is predicted, and
this is confirmed by
experiment. The detriment in the noise figure previously associated with counterpropagating
schemes for only
small-signal conditions is shown to be similar for strong signals, corresponding to a power-amplifier
operating
regime. In addition, a correction to the noise figure is given for the limit of a low-gain amplifier.

The erbium-doped silica fiber amplifier1' 2 is receiving ASE. If weassume rapid relaxation from pump bands
much interest as an optical amplifier in the third opti- into the metastable level and uniform dopant and field
cal telecommunications window near 1.55 gm. Ad- distributions across the fiber core, at a given axial
vantages of erbium-doped fiber amplifiers include position z in the fiber these can be written as
high gain that is polarization independent, the ab-
sence of Fresnel reflections, and quantum-limited
noise behavior.3 In addition, the recent demonstra- dN2(Z) = Wp(z) [Ntot - N 2 (Z)]
tions of excellent performance and near-100% quan-
tum efficiency when pumped at wavelengths free from
pump excited-state absorption 4 (ESA) indicate that - WS(z)[(1+ a)N2(z) - aNtotl- N 2 (z) (1)
compact and practical devices will soon become a real- T21
ity.
In this Letter we present results of a full strong- dP +(z)
signal numerical analysis of the amplifier to describe = - (Z)abs[Ntt- N 2 (Z)]
the noise and gain characteristics for both copropagat- dz
ing and counterpropagating signal pump schemes.
Previous research has analyzed the problem on the - P+(Z)1YESAN2(Z) (2)
basis of negligible amplified spontaneous emission
5
(ASE) and ESA or for small signals only in the coun-
terpropagating scheme.6 A comprehensive treatment
dP' () =z +-(Z)z(Z),
(3)
of the effect of modal overlaps and ESA on small- dz
signal gain has been given by Armitage,7 and the spec- dPf'-(z) = t.t(z)hvAvy(z)
+ P`Z-()
tral evolution of ASE has been given by Desurvire and (4)
dz
Simpson.8 In this research we make approximations
with regard to modal overlaps and ASE in order to local gain Y(z)= 7Mo21[(1
+ a)N2(Z) - aNtj. (5)
simplify the approach, although without sacrificing
the limiting effects of either. We show that under Here N 2 and Ntot are the metastable-level population
strong-signal conditions the noise detriment associat- density and the total erbium dopant concentration,
ed with the counterpropagating scheme is similar to respectively, Wpis the pump rate, W, is the stimulated
that of the small-signal case, and that for amplifiers emission rate, and Tf is the metastable-level lifetime,
operating in the low-gain limit the noise figure goes to where rf has been measured as (12 4 0.2) msec for the
unity, as expected from physical considerations. two germanosilicate fibers used here. In equilibrium,
However, we also show that a difference in strong- i.e., dN 2(z)/dt = 0, Eq. (1) can be explicitly written as
signal gain is expected between the two possible pump
schemes in the presence of pump ESA, with the coun- N2 (z) = Ntot[Wp(z) +aW8 (z)J
WP(z) (6)
terpropagating scheme showing the greatest gain. + (1 + a))W
8 (z) + l-f
This result is shown to agree with experimental mea-
surements. The pump and stimulated emission rates, Wp(z) and
The behavior of the erbium amplifier, in common W,(z), are given by
with other rare-earth-doped fiber devices, can be de-
scribed in terms of rate equations for the population- -Ws(z) =
inversion density, the pump field, the signal field, and hv) (7)
hv,,a 21's

0146-9592/89/191062-03$2.00/0 1989 Optical Society of America

 October 1, 1989 / Vol. 14, No. 19 / OPTICS LETTERS
W,(z) = aabsflp' (8)
where a is the core area of the fiber and 1p and ns are
the proportion of the pump and signal powers propa-
gated within the fiber core. Pp, Ps, and Pfl are the
pump, signal, and ASE powers, respectively (the plus
sign indicates copropagating and the minus sign indi-
cates counterpropagating relative to the pump direc-
tion), and cabs, IYESA,and c21 are the pump absorption,
pump ESA, and stimulated emission cross sections,
respectively. The term a is given by the ratio a,2/a21
at the signal wavelength. Although we are aware of an
analysis that gives a12/(r2l < 1 for aluminosilicate fi-
8
bers using a Fuchtbauer-Ladenburg relation, for this
fiber type by comparison of the small-signal um-
pumped attenuation with the small-signal gain under
conditions of heavy pumping at 980 nm we find that a
- 1.25 i1 0.1. In the expression describing ASE, hp
and Av are the photon energy and optical bandwidth of
the ASE, respectively, and the term ,u is defined as
N2 (z)I[(1 + a)N 2(z) - aNtot]. The ASE bandwidth is
taken to be 2 nm on the basis that ASE-induced satu-
ration is experimentally observed to occur when the
ASE spectrum has narrowed to approximately 2 nm.
Hence, although this value will give ASE powers that
are too low at low gains, the important limit of ASE-
induced gain saturation is accounted for. These equa-
tions can be integrated numerically for arbitrary sig-
nal powers in the copropagating and counterpropagat-
ing configurations, and the solutions can be iterated to
give a self-consistent condition for the signal, the ASE,
and the pump fields.
Numerical solution of the above equations was per-
formed for 665-nm pumping of a 0.21-N.A., 1050-nm
cutoff germanosilicate fiber doped with -8.8 X 1018
cm- 3
Er 3+
ions (400 parts in 106). A value of 6.5 X
0
10-21 cm 2
(Ref. 9) was used for 21, and abs was in-
ferred from spectral attenuation data of a highly mul-
timode fiber to be approximately half this value. At a
665-nm pump wavelength the ratio 6ESA/cabs has been
measured to be -0.2.10 The overlap of pump radia-
tion with the core was assumed to be complete owing
to the multimode nature of the pump, and the signal
overlap with the core (as) was taken to be 0.4. Figure 1
shows the predicted variation of gain with the fiber
length for 0.1-MWand 100-AWsignal inputs at 1.536
Aim,for 100 mW of pump power, in the copropagating
and counterpropagating pump schemes. Superim-
posed onto the theoretical3 curves are experimental
values previously reported, which were obtained with
a fiber characterized by the above parameters. One
interesting feature of the theoretical curves is the
predicated difference in gain in the saturated regime
between the copropagating and counterpropagating
pump schemes. At the position of maximum gain the
counterpropagating scheme is expected to show more
gain, and this is confirmed in the experimental data
for the 100-,gWsignal input. This feature is believed
to result from a reduction of pump ESA owing to
reduced inversion in the pump input end at high signal
powers with the counterpropagating scheme. Using
the overlap of ais= 0.4 (slightly lower than the propor-
1063
tion of the power propagated in the core, 0.62, at this
fiber V value 1 ) the maximum gain for the two cases is
predicted. However, the saturation region is experi-
mentally observed to be more spread out than that
predicted by the simple model. We believe this may
be due to the interplay of the field and dopant distri-
butions in the fiber, which have been generalized in
this treatment.
In a similar way, numerical calculations were per-
formed for the 980-nm pump wavelength that has re-
cently4 been shown to be highly efficient owing to the
0 Experimental data were ob-
absence of pump ESA.1
tained with a different fiber, although it had nominal-
ly the same core composition as the previous fiber.
The parameters that were different from the 665-nm
pump case were N.A. = 0.16, a cutoff 3 of 975 nm, an
Er 3 + concentration of 5.1 X 1018 cm- , (abs = 4.9 X
10-21 cm 2 , #IESA/aabs = 0, and a pump power of 15 mW.
The pump overlap with the dopant distribution was
assumed to be 83%,which corresponds to the propor-
tion of the fundamental mode propagated in the corell
and the signal overlap of 0.37. Note that this signal
overlap is in the same ratio (0.65) to the signal power
propagated in the corell as with previous fiber. Fig-
ure 2 shows the experimental and theoretical data for
the 980-nm pump wavelength obtained with the above
fiber. A good match is again observed between theory
and experiment. Note that in this case the theoretical
50
40
Gain 30
(dB) 20
10
0
0 1 2 3 4 5 6 7 8
Fiber length (i)
Fig. 1. Theoretical (curves) and experimental (points) gain
versus the fiber length for 100mW of pump power at 665nm.
Copropagating signal inputs were 0.1 pW (open circles) and
100 AW (filled circles) at 1.536 Am; counterpropagating sig-
nal inputs were 0.1 ,W (crosses) and 100 iW (diamonds) at
1.536 ,m.
40
30
Gain 20
(dB)
10
0
0 2 4 6 8 10 12
Length (im)
Fig. 2. Theoretical (curves) and experimental (points) gain
versus the fiber length for 15 mW of pump power at 980 nm
for copropagating and counterpropagating 0.1-MW(circles)
and 100-AW(squares) signal inputs at 1.536,m.
 1064 OPTICS LETTERS / Vol. 14, No. 19 / October 1, 1989
12
10 counter-prop
NF - 6 large signal
(dB1) 6
4 The higher signal level is also seen to increase the
co-prop
2 copropagating noise figure owing to a reduction of
/ small signal
0 1 2 3 4 5 6 7 8 higher signal level the degradation in the signal-to-
12
10
8 - large signal counter-prop
NF
(d 1 6)
4 shansky,6 and it follows from simple physical reason-
2 ~~~~~~~co-proping that the noise figure should go to unity under such
small signal
0II
o 2 4 6 8 10 12 14
Fiber length (m)
Fig. 3. Theoretical noise figure (NF) versus the fiber length The results show that in the presence of ESA there
for (a) 100 mW of pump power at 665 nm and (b) 15 mW of
pump power at 980 nm. The signal inputs are 0.1 AtW
(small) and 100 /W (large) at 1.536 /um.
data for copropagating and counterpropagating signal
pump schemes were derived to be the same within the
errors of the numerical model and are indistinguish-
able in Fig. 2. Hence we conclude that the presence of
pump ESA in the case of 665-nm pumping is responsi-
ble for a difference in the strong-signal gain.
It has been shown that the noise induced into an
optical signal by the erbium amplifier is dominated by
a combination of signal-spontaneous beat noise and
spontaneous-spontaneous beat noise.3 Since optical
filtering can potentially be used to reduce the sponta-
neous-spontaneous contribution, the limiting noise
source is likely to be signal-spontaneous beat noise.
A convenient way of modeling the signal-spontaneous
beat noise detriment of the amplifier is to plot the
noise figure (NF) of the amplifier defined in a manner
similar to that in Ref. 6, NF = [2,geff(G- 1)/G + 11G],
against the amplifier length, where G is the single-pass
gain of the amplifier (defined as the exponential of the
integral of the local gain over the fiber length) and Meff
is defined as
eff 2hvAv(G-1) ()
In physical terms Jeff relates to M(z), which is defined
above, in the region of the amplifier where (G - 1)IG is
considerably less than unity. In practice, this corre-
sponds to the signal input end of the fiber. Note that
eff(G - 1) is equivalent to Np in Ref. 6. Figure 3
shows the predicted noise figure for 0.1-AWand 100-
,W signals pumping at (a) 665 nm and (b) 980 nm.
The small-signal results in both cases agree closely
with the data of Olshansky,6 indicating a degradation
in the signal-to-noise ratio of -6 dB at the point of
maximum gain in the counterpropagating case and
just over the 3-dB quantum limit for the copropagat-
ing case. The copropagating noise figure for 980-nm
pumping is slightly higher than that of 665-nm pump-
ing because of the lower initial inversion that results
(a) from a substantially lower pump power in this case.
inversion at the signal input end of the fiber. At the
noise ratio is again seen to be -6 dB for the counter-
propagating case at the optimum length both with and
without ESA. Note, however, that the noise figure is
shown to go to unity at vanishingly small amplifier
(b) lengths. This low-gain limit was not treated by 01-
conditions.
We have shown the implementation of a numerical
model to describe the behavior of the erbium amplifier
in the presence of ESA and for arbitrary signal inputs.
exists a difference in strong-signal gain between co-
propagating
and counterpropagating pump schemes,
with the counterpropagating scheme being favorable.
This result is validated by experimental data. In ad-
dition, we have shown that in the strong-signal limit
the degradation in the signal-to-noise ratio at the fiber
length giving maximum gain is expected to be similar
to the small-signal case, both with and without ASE.
A correction to the noise figure for low-gain (<10-dB)
amplifiers has also been given.
R. I. Laming acknowledges support by Pirelli Gen-
eral plc.
References
1. R. J. Mears, L. Reekie, I. M. Jauncey, and D. N. Payne,
Electron. Lett. 23, 1027 (1987).
2. E. Desurvire, J. R. Simpson, and P. C. Becker, Opt. Lett.
12,888 (1987).
3. R. I. Laming, P. R. Morkel, D. N. Payne, and L. Reekie,
in Proceedings of European Conference on Optical
Communications (Institution of Electrical Engineers,
London, 1988), p. 54.
4. R. I. Laming, L. Reekie, D. N. Payne, P. L. Scrivener, F.
Fontana, and A. Righetti, in Proceedings of European
Conference on Optical Communication (Institution of
Electrical Engineers, London, 1988),Pt. 2, p. 25.
5. M. Montecchi, A. Mecozzi, and M. Tamburrini, present-
ed at the European Conference on Optics and Optical
Systems and Applications, Birmingham, UK, 1988.
6. R. Olshansky, Electron. Lett. 24, 1363 (1988).
7. J. R. Armitage, Appl. Opt. 27,4831 (1988).
8. E. Desurvire and J. R. Simpson, IEEE J. Lightwave
Technol. LT-7, 835 (1989).
9. V. P. Gabontsev, S. M. Matitsiu, A. A. Iseneev, and V. B.
Kravchenko, Opt. Laser Technol. (August 1982),p. 189.
10. R. I. Laming, S. B. Poole, and E. Tarbox, Opt. Lett. 13,
1084 (1988).
11. A. W. Snyder and J. D. Love, Optical Waveguide The-
ory (Chapman & Hall, London, 1983),p. 314.